Abstract:For stochastic Volterra equations driven by standard Brownian and with singular kernels $K(u)=u^{H-\frac{1}{2}}/\Gamma(H+1/2), H\in (0,1/2)$, it is known that the Milstein scheme has a convergence rate of $n^{-2H}$. In this paper, we show that this rate is optimal. Moreover, we show that the error normalized by $n^{-2H}$ converge stably in law to the (nonzero) solution of a certain linear Volterra equation of random coefficients with the same fractional kernel.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the limit error distribution of the Milstein numerical scheme for stochastic Volterra equations (SVE) with singular kernels. Specifically, the paper focuses on the following two issues:
1. **Is the convergence rate optimal?** It is known that for the stochastic Volterra equation driven by the standard Brownian motion and with the singular kernel \(K(u)=\frac{u^{H - 1/2}}{\Gamma(H+ 1/2)}\), the convergence rate of the Milstein scheme is \(n^{-2H}\). The paper aims to prove that this convergence rate is optimal.
2. **Limit distribution of the normalized error**: The paper further studies whether, when the error is normalized by \(n^{-2H}\), it stably converges in distribution to a non - zero solution of a certain linear stochastic Volterra equation. This particular linear stochastic Volterra equation has the same fractional kernel and random coefficients.
### Main Results
The main results of the paper show that the answers to the above questions are affirmative. Specifically, let \(U^n_t=n^{2H}(X_t - X^n_t)\), where \(X_t\) is the solution of the original stochastic Volterra equation, and \(X^n_t\) is the numerical solution obtained using the Milstein scheme. Then the process \(U^n_t\) stably converges in distribution in the space \(C_0^{H-\epsilon}([0, T], \mathbb{R}^d)\) to a unique solution process \(U_t=(U^1_t,\dots,U^d_t)^T\), which satisfies the following linear stochastic Volterra equation:
\[
U^i_t=\sum_{k = 1}^d\int_0^tK(t - s)\partial_kb_i(X_s)U^k_sds+\sum_{j = 1}^m\sum_{k = 1}^d\int_0^tK(t - s)\partial_k\sigma^j_i(X_s)U^k_sdW^j_s
\]
\[
+C_M\sum_{j = 1}^m\sum_{i'= 1}^m\sum_{l = 1}^d\sum_{l'= 1}^d\int_0^tK(t - s)\partial_k\sigma^j_i(X_s)\sigma^l_i(X_s)\partial_l\sigma^{i'}_k(X_s)dB^{i,i',j}_s
\]
where \(B\) is an \(m^3\)-dimensional standard Brownian motion independent of the original Brownian motion \(W\), and \(C_M\) is a constant related to the Milstein scheme.
### Methodology
To prove these results, the authors applied the fractional - order stochastic calculus and the limit theorems of fractional - part integrals. The main technical difficulty lies in giving the exact expressions of the constants appearing in the limit equations and overcoming these difficulties through a series of lemmas and technical means.